Font Size: a A A

Research On Load Balance Strategy Of Web Server Based On Nginx

Posted on:2020-03-03Degree:MasterType:Thesis
Country:ChinaCandidate:Y GaoFull Text:PDF
GTID:2428330572469897Subject:Electronics and Communications Engineering
Abstract/Summary:PDF Full Text Request
In a large number of concurrent requests,multi-server clustering technology can effectively process requests and improve the overall processing power of the server.As a core,load balancing has a decisive role in properly distributing requests and improving the overall performance of the server.At present,in the technology of solving concurrent web servers,Nginx has the characteristics of strong concurrent processing and good scalability,which has attracted a lot of attention,research and application.Based on the research of Nginx load balancing algorithm,this paper proposes two feasible and effective improved algorithms.The detailed research work and gains are as follows:1.For the Nginx default load balancing strategy,the load of the Web server cluster has not been fully covered by the situation that the service request is always changing,and the MRD strategy is innovatively proposed.To alleviate the processing pressure of load balancing,the back-end Web server load information can be reported less by effectively controlling the load change rate and the load reporting period.By comparing the MRD strategy with Nginx weighted polling and "rt_dyn_load algorithm",the results show that when processing the request,the strategy can first determine the load status of the cluster,and then select the best according to the pressure situation of each load state.The background server,in turn,effectively reduces the average latency of the Web server cluster,resulting in a significant increase in the average system throughput.2.Aiming at the fact that the MRD strategy adopts weighted polling without exceeding the threshold,and the weighted polling has many drawbacks,a dynamic load balancing algorithm based on polling(DWRS)is proposed by analyzing the Nginx source code.In addition to real-time collection and processing of client load requests,the algorithm can effectively classify these requests and convert them into opposite weights.Finally,the requests are allocated according to the weights.In the Linux system,the algorithm and Nginx built three load balancing strategies are tested by building heterogeneous clusters.The experimental results show that the algorithm is better than the Nginx built-in in terms of average response time and actual concurrent number.Load balancing strategy.3.Compare the DWRS proposed by this paper with the MRD strategy.The experimental data shows that DWRS has an absolute advantage in the pursuit of processing speed,but MRD is slightly better if the average throughput is pursued.In short,the two algorithms proposed in this paper have their own merits.
Keywords/Search Tags:load balance, minimum residual, Nginx, DWRS
PDF Full Text Request
Related items